Uncertainty management in radioactive waste repository site assessment

Abstract

The problem of performance assessment of a site to serve as a repository for the final disposal of radioactive waste involves different types of uncertainties. Their main sources include the large temporal and spatial considerations over which safety of the system has to be ensured, our inability to completely understand and describe a very complex structure such as the repository system, lack of precision in the measured information, etc. These issues underlie most of the problems faced when rigid probabilistic approaches are used. Nevertheless a framework that would allow for an optimal aggregation of the available knowledge and an efficient management of the various types of uncertainty involved is needed. In this work a knowledge-based modelling of the repository selection process which through a consequence analysis evaluates the potential impact that hypothetical scenarios will have on a candidate site is proposed. The model is organised around a hierarchical structure, linking the relations between the scenarios and the possible events and processes that characterise them, as well as the affected site parameters. The scheme provides for both crisp and fuzzy parameter values and uses fuzzy semantic unification and evidential support logic inference mechanisms. It is implemented using the A.I. language FRIL and the interaction with the user is performed through a windows interface.